Earth observation models with spatial validation
Land cover, building and damage models are trained and tested on spatially separated blocks, so reported accuracy reflects new places rather than the neighbours of training pixels.
The Global Centre for Risk and Innovation (GCRI)
Artificial intelligence and machine learning extend detection, classification and forecasting across large archives, and agents extend them into action. In this field the method is also a source of risk: models that fail when conditions shift, outputs that sound certain when they are not, and tools that turn an error into an effect.
Machine learning model governance is the set of practices that decide which models may be used, for what and on what evidence. It covers validation designed for the data, such as spatial cross-validation for earth observation, model cards and promotion records, red-team findings and agents limited to declared scopes with receipts for their actions.
The working screen behind this method: its structure, terms and states. Names are placeholders and values illustrative.
Foundry · AI and machine learning at Nexus… · Run R-6794
| All | 0.83 | 0.02 | 5712 | Pass |
| Region A | 0.74 | 0.05 | 1098 | Review |
| Region B | 0.76 | 0.05 | 458 | Review |
| Rare class | 0.64 | 0.07 | 160 | Below bar |
Artificial intelligence and machine learning develop algorithms that learn patterns from data to classify, predict or generate outputs. They are increasingly used for hazard detection, forecasting and decision support, and are themselves a source of risk. In data and compute and methods and modelling, machine learning can scale detection and forecasting across interconnected systems. Its own failure modes must be weighed when automated outputs inform consequential decisions. The canonical description is on the STEM Commons.
Land cover, building and damage models are trained and tested on spatially separated blocks, so reported accuracy reflects new places rather than the neighbours of training pixels.
Each model version carries its intended use, training data, evaluation by region and subgroup, known limits and red-team results, and is promoted only with a recorded review.
Agents run analyses with a declared tool set, data classes, budget and kill-switch enforced at the policy gate, and every action produces a receipt for a reviewer.
Each term as this page uses it. Where a standard-setting body defines the term, the wording follows that body.
| Term | Definition |
|---|---|
| Spatial cross-validation | Validation that holds out whole regions, so a model is tested on places it has not seen. |
| Model promotion | The recorded decision to move a model from testing into use, with its evidence. |
| Out-of-distribution test set | Data deliberately unlike the training data, used to see how a model fails. |
Urban demand centres and the hubs that anchor digital and data infrastructure, shown for Middle East and North Africa as an example geography for this subject.
Data from the source register feeds the methods; calibration and an independent test gate the estimate and its limits; the decision stays with the competent authority.
Fig. 01Nodes carry their type, metrics and state; connectors show what feeds what; the gates are the validation steps; the timeline is one application. Design of the method's use, not a record of a run.
Datasets and records used by artificial intelligence and machine learning on this platform, with the role each plays and the label its values carry. Sources are listed in the source register.
| Data | Role | Label |
|---|---|---|
| Labelled training sets | Supervised learning | REPORTED |
| Satellite image archives | Inputs to classification | OBSERVED |
| Held-out and out-of-distribution test sets | Evaluation | OBSERVED |
| Evaluation and red-team records | Promotion decisions | REPORTED |
| Datasheets for training data | Origin, consent and licence | DECLARED |
Models are evaluated on held-out data that respects spatial, temporal and group structure. Robustness, calibration and fairness are tested alongside accuracy.
| Check | How it is done |
|---|---|
| Spatial and temporal hold-out | Test data are separated in space and time from training data to prevent leakage |
| Probability calibration | Predicted probabilities are compared with observed frequencies on held-out data |
| Drift monitoring | Input and output distributions in use are compared with those at validation, and divergence triggers review |
| Adversarial testing | Models and agents are tested against prompt injection and misuse before promotion |
Performance degrades when deployment conditions differ from training data. Models can encode biases and fail without clear warning on rare events.
Artificial Intelligence and Autonomous Systems stewards artificial intelligence and machine learning for the Consortium and reviews how it is applied across platforms.
The current service level is R0: information and enquiries. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.